Modern communications and data networks are comprised of nodes that transport data through the network. The nodes may include routers, switches, bridges, or combinations thereof that transport the individual data packets or frames through the network. Some networks may offer data services that forward data frames from one node to another node across the network without using pre-configured routes on the intermediate nodes. Other networks may forward the data frames from one node to another node across the network along pre-configured or pre-established paths. In some networks, the nodes may create Ethernet-Local Area Network (E-LAN) services, where traffic that corresponds to different services may be transported along different sub-networks, e.g. by different subsets of nodes. For example, the E-LAN services may comprise Institute of Electrical and Electronics Engineers (IEEE) 802.1aq network services or Virtual Private LAN Services (VPLS).
For some premium data services, source nodes (e.g., edge nodes such as a Customer Edge node, a Service Provider Ingress node, a multiprotocol label switching (MPLS) label switched path (LSP) source node, or a provider backbone bridging (PBB)-traffic engineering (TE) traffic engineered service instance (TESI) source node) may have multiple paths, which may be purchased from different service providers, to their corresponding destination nodes. For such premium services, the optimal path from each service provider is desired to achieve the best delivery of their traffic. However, condition changes (e.g., due to congestion on transit nodes, microwave transport bandwidth being impaired due to weather, or downstream hop port changes) along the pre-established path (e.g., an LSP, a PBB-TE path, or other transport path) may necessitate changes over time.
Various techniques exist to signal path impairment; however, the known techniques are either inefficient or otherwise impractical. For example, resource reservation protocol (RSVP)-TE enables individual links to advertise its available bandwidth to all the nodes in the routing domain, but RSVP-TE is not possible if the path between the source node and the destination node traverse multiple routing domains. As another example, Internet Protocol (IP) Explicit Congestion Notification (ECN) (Internet Engineering Task Force (IETF) Request for Comments (RFC) 3168) and Transmission Control Protocol (TCP) ECN (RFC 5562) describe randomly marking on data frames in transit when congestion occurs in the middle of the network. However, some transit nodes do not detect IP frames and some egress nodes (e.g., MPLS LSPs) do not terminate all of the IP frames. Also, MPLS ECN (RFC 5129) specifies a mechanism for transit nodes to mark experimental (EXP) bits when congestion happens. However, many deployed MPLS networks already use EXP bits to mark priority, and thus the ECN (RFC 5129) mechanism cannot be implemented in those networks.
One issue with ECN (e.g., IP ECN, TCP ECN, and MPLS ECN) is that the congestion marking does not occur until congestion happens. Accordingly, when a transit link bandwidth is reduced (e.g., when a microwave transport link's bandwidth is reduced due to weather), the queue on the transit node can quickly build up. Even if the ECN (IP/TCP/MPLS ECN) scheme is used, by the time the egress node recognizes the congestion and notifies the source node, the queue on the transit node may already overflow resulting in lost packets. IEEE 802.1au specifies a sophisticated mechanism and algorithm to enable an intermediate node to indicate the congestion level to the source node. However, the quantized feedback algorithm is complicated and the source node may not need the quantity and type of information provided.
Another issue with ECN and IEEE 802.1au is that the source node may not even be able to do anything with the congestion notification. For example, the source node may only have one link out to the destination or may not have sufficient capability to switch traffic. In these circumstances, the source nodes do not need to know the congestion condition in the middle of the network. Therefore, the marking operations performed by intermediate nodes upon congestion or impairment are wasted.